Merge branch 'ASUS-AICS:master' into pipeline_tutorial

Author: Gordon119

example_config/EUR-Lex-57k/README.md

@@ -11,4 +11,4 @@ The best model is selected by **RP@5** on validation set then evaluated on test
 |     Kim-CNN     | [Chen et al. 2022](https://www.csie.ntu.edu.tw/~cjlin/papers/xmlcnn/xml_cnn_study.pdf) |     20.0767     |     69.9073     |     90.7833     |     66.8733     |     77.5347     |     80.3488     | [Cfg](./kim_cnn.yml) | 30 mins |
 |     CNN-LWAN     | [Mullenbach et al. 2018](https://aclanthology.org/N18-1100/) |     26.3245     |     71.6865     |     89.8167     |     67.3100     |     78.0261     |     80.5446     | [Cfg](./cnn_lwan.yml) | 2 hrs |
 |     BiGRU-LWAN     | [Chalkidis et al. 2019](https://aclanthology.org/P19-1636/) |     25.4260     |     71.3199     |     90.7333     |     67.0800     |     77.7506     |     80.5999     | [Cfg](./bigru_lwan.yml) | 3 hrs 20 mins |
-|     BERT-BASE-LWAN     | [Chalkidis et al. 2020](http://aclanthology.lst.uni-saarland.de/2020.emnlp-main.607/) |     26.1981     |     72.7849     |     89.9833     |     67.8533     |     78.4742     |     80.9834     | [Cfg](./bert_lwan.yml) | 9 hrs |
+|     BERT-BASE-LWAN     | [Chalkidis et al. 2020](https://aclanthology.org/2020.emnlp-main.607/) |     26.1981     |     72.7849     |     89.9833     |     67.8533     |     78.4742     |     80.9834     | [Cfg](./bert_lwan.yml) | 9 hrs |

example_config/EUR-Lex-57k/bert_lwan_tune.yml

@@ -4,7 +4,7 @@ val_file: data/EUR-Lex-57k/valid.txt
 test_file: data/EUR-Lex-57k/test.txt
 data_name: EUR-Lex-57k
 min_vocab_freq: 1
-max_seq_length: 500
+max_seq_length: 512
 include_test_labels: true
 remove_no_label_data: false
 add_special_tokens: false

search_params.py

@@ -171,13 +171,11 @@ def prepare_retrain_config(best_config, best_log_dir, retrain):
         best_config.merge_train_val = False
 
 
-def load_static_data(config, merge_train_val=False):
+def load_static_data(config):
     """Preload static data once for multiple trials.
 
     Args:
         config (AttributeDict): Config of the experiment.
-        merge_train_val (bool, optional): Whether to merge the training and validation data.
-            Defaults to False.
 
     Returns:
         dict: A dict of static data containing datasets, classes, and word_dict.
@@ -187,7 +185,7 @@ def load_static_data(config, merge_train_val=False):
         test_data=config.test_file,
         val_data=config.val_file,
         val_size=config.val_size,
-        merge_train_val=merge_train_val,
+        merge_train_val=config.merge_train_val,
         tokenize_text="lm_weight" not in config.network_config,
         remove_no_label_data=config.remove_no_label_data,
     )
@@ -231,7 +229,7 @@ def retrain_best_model(exp_name, best_config, best_log_dir, retrain):
     with open(os.path.join(checkpoint_dir, "params.yml"), "w") as fp:
         yaml.dump(dict(best_config), fp)
 
-    data = load_static_data(best_config, merge_train_val=best_config.merge_train_val)
+    data = load_static_data(best_config)
 
     if retrain:
         logging.info(f"Re-training with best config: \n{best_config}")